English | 简体中文(Simplified Chinese)
FuncGenFoil, short for Function-Space Generated Airfoil, is a method for generating airfoils using generative models in function space, such as diffusion models or flow models. This library provides a framework to demonstrate the power of generative models in design and optimization.
Demonstration of airfoil generation model training and inference in function space.
Demonstration of airfoil editing model training and inference in function space.
Please install from source:
git clone https://github.com/zjowowen/FuncGenFoil.git
cd FuncGenFoil
pip install -e .
Here is an example of how to train a airfoil generative flow model in function space.
Download dataset from here and save it in the current directory.
To train the model without conditional information:
python examples/train_unconditional_airfoil_generation.py
To train the model with conditional information:
python examples/train_conditional_airfoil_generation.py
To train the model for airfoil editing:
python examples/train_airfoil_editing.py
To evaluate the model for airfoil generation with super-resolution:
python examples/eval_airfoil_generation_super_resolution.py
If you find this work useful, please consider citing:
@misc{zhang2025funcgenfoilairfoilgenerationediting,
title={FuncGenFoil: Airfoil Generation and Editing Model in Function Space},
author={Jinouwen Zhang and Junjie Ren and Aobo Yang and Yan Lu and Lu Chen and Hairun Xie and Jing Wang and Miao Zhang and Wanli Ouyang and Shixiang Tang},
year={2025},
eprint={2502.10712},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2502.10712},
}
FuncGenFoil is licensed under the Apache License 2.0. See LICENSE for more details.